The purpose of this investigation is to improve the precision of early identification screening for reading failure and to determine the extent to which secondary prevention modifies its probability. The emotional and economic sequelae of reading failure in a highly literate society represent a public health concern. Despite the unprecedented national attention given to the problem of reading failure, a model that accurately identifies who will and who will not exhibit early reading failure does not exist. The number of potential marker variables are legion but have yet to be coordinated into a cohesive system that is theoretically robust and systematically tested. In addition, the duration of research-based reading intervention and its long term effects on at-risk children are rarely studied. Two research questions are designed to address the limitations of past research: (a) Is the precision of early identification screening improved by incorporating fluency and growth measures;and (b) Does the probability of reading failure defined by the early identification screen interact with instructional duration to modify the risk of reading failure both immediately and longitudinally? A novel aspect of the design is defining risk of reading failure continuously as probability level rather than dichotomously. The research plan incorporates a multiple cohort design in which three successive cohorts of first grade children are selected (N=250-300 across years). In Year 1, measures tested for the early identification screen include accuracy and fluency measures collected as point estimates (static measurement) and across time (growth measurement). We hypothesize that the lack of attention to fluency and growth has hampered past early identification efforts. Multiple regression and dominance analysis are used to assess impact on reading measured continuously and to reduce the variable set. Logistic regression is used to develop the best fitting algorithm to predict reading failure and produce probabilities. This algorithm is applied in Years 2 and 3 to identify children with higher probabilities of failure (N=180). These children are assigned to a no treatment control or to intervention that lasts 10 weeks or 16 weeks. Duration is varied to understand what is gained and lost by the intervention lengths typically used in reading intervention research. All three cohorts are followed through third grade with the diagnostic battery.